# machine-learning
- p
Pikachu ✓
12/15/2022, 8:05 PMYeah, I know - h
hecko
12/15/2022, 8:10 PMit's possible to generate way longer snippets, though not really train on them - h
hecko
12/15/2022, 8:10 PMidk how consistent it is because i've only ever done img2img - h
hecko
12/15/2022, 8:11 PMalso someone figure out how to upscale from 750 hertz to listenable audio, which could help with generating longer snippets - h
hecko
12/15/2022, 8:12 PMhttps://cdn.discordapp.com/attachments/1034910228543713331/1049785168065462352/sample_0_downsampled.wav https://cdn.discordapp.com/attachments/1034910228543713331/1049785168392630363/sample_1_upsampled.wav https://cdn.discordapp.com/attachments/1034910228543713331/1049786959284936785/sample_1_true.wav - p
Pikachu ✓
12/15/2022, 8:34 PMWho did this - p
Pikachu ✓
12/15/2022, 8:35 PMPlus can we use it for 44 khz - p
Pikachu ✓
12/15/2022, 8:35 PMAnd ai voices - h
hecko
12/15/2022, 8:35 PMyes, and idk - p
Pikachu ✓
12/15/2022, 8:39 PMBruh what - p
Pikachu ✓
12/15/2022, 8:39 PMYou don't know the source? - h
hecko
12/15/2022, 8:41 PMidk if we can use it for ai voices - h
hecko
12/15/2022, 8:41 PMlook up
on harmonai discord750 - p
Pikachu ✓
12/15/2022, 8:42 PMBru - p
Pikachu ✓
12/15/2022, 8:42 PMIt's harmonak - p
Pikachu ✓
12/15/2022, 8:51 PMAlso @hecko question - h
hecko
12/15/2022, 8:51 PM? - p
Pikachu ✓
12/15/2022, 8:52 PMIs it actually possible to dreambooth music spectrogram - h
hecko
12/15/2022, 8:52 PMyes - h
hecko
12/15/2022, 8:52 PMriffusion is just a finetune - p
Pikachu ✓
12/15/2022, 8:52 PMOmg - h
hecko
12/15/2022, 8:52 PMand dreambooth is just finetuning but different in ways i'm unfamiliar with - p
Pikachu ✓
12/15/2022, 8:53 PMCan I just do it with Ed Sheeran music - h
hecko
12/15/2022, 8:53 PMyes - p
Pikachu ✓
12/15/2022, 8:53 PMBru - h
hecko
12/15/2022, 8:53 PMbut the code for making the spectrograms isn't out yet - p
Pikachu ✓
12/15/2022, 8:53 PMThey need to release it - h
hecko
12/15/2022, 8:54 PM-actually no wait there is